How Do You Audit Whether AI Actually Cites Your Brand?

A sister property of ours scored a 45% AI-citation rate overall but only 35% on ChatGPT (chudi.dev internal panel, 2026-06-12), and the ChatGPT misses were not random: the engine kept quoting a mirror domain instead of the canonical site. We only knew because we tested it. Most brands never do, which is exactly the problem an AI visibility audit solves.

An AI visibility audit tests whether AI answer engines (ChatGPT, Perplexity, Gemini, Google AI Overviews) cite your brand for your buyers' real questions, measures your citation rate per engine, identifies which competitors are cited in your place, and surfaces the on-site gaps (crawler access, schema, answer structure) that decide who gets quoted.

Here is the failure mode almost everyone starts with: they assume that ranking well on Google means AI engines see them too. They do not. The engines build answers from sources they can parse and trust, and a page that ranks #3 on Google can be invisible inside ChatGPT. This is the same gap the 0% ChatGPT citation trap describes, and this post shows you the 6 checks that make it measurable.

What is an AI visibility audit?#

An AI visibility audit is a structured test of whether AI engines name your brand when buyers ask about your category. It produces three numbers: your citation rate per engine, the list of competitors cited instead of you, and a ranked set of fixes. Unlike an SEO audit, it measures presence inside generated answers, not position on a results page.

The reason this matters now is commercial, not academic. When we pulled live demand (DataForSEO, US/en, 2026-06-21), "ai visibility audit" showed only 90 monthly searches but a $39.79 CPC at keyword difficulty 6. That combination, low volume and high cost-per-click, is the signature of an early, high-intent buyer market: few people search it, but the ones who do are ready to pay.

Why can't Google Analytics tell you this?#

Google Analytics cannot tell you whether AI cites you because AI answers often produce zero clicks. When ChatGPT summarizes your content without linking, or Perplexity cites you in a footnote the user never clicks, no session ever reaches your site. The influence happened entirely off your property, so your analytics show nothing.

That is the core blind spot. 68% of US Google searches now end without a click (SparkToro + Similarweb, early 2026), and AI Overviews push that even higher. Your analytics measure the clicks you still get; they are structurally blind to the answers that replaced the clicks you lost. An AI visibility audit measures that invisible layer directly by querying the engines themselves.

How do you run an AI visibility audit? The 6 checks#

You run an AI visibility audit by testing your buyer questions across each engine, scoring citations, mapping competitors, and inspecting your crawlability and content structure. Below are the 6 checks, in order, each producing a concrete artifact you can act on.

Check 1: Define your buyer-question set#

Write the 15 to 20 questions a real buyer asks an AI assistant before purchasing in your category. Phrase them as natural questions, not keywords. This set is the spine of the entire audit, so be honest: include the "best tool for X" and "X vs Y" questions where you most want to appear.

Check 2: Test each question across all four engines#

Run every question through ChatGPT, Perplexity, Gemini, and a Google AI Overview query, in a clean non-personalized session. For each result, record whether your domain was cited, the exact URL cited, and which competitor domains appeared. This is the raw evidence; everything downstream derives from it.

Check 3: Score your citation rate per engine#

Calculate citations divided by questions, separately for each engine. Per-engine matters more than the blended average: a 45% overall rate can hide a 35% ChatGPT rate, and ChatGPT is often where your buyers actually are. Never accept a single composite "visibility score" that collapses this detail, because it hides the engine where you are losing.

Check 4: Map the competitors cited in your place#

For every miss, log who got cited instead. Two or three domains usually own most of your category's AI answers, and that set is frequently different from your Google competitors. This list tells you who the engines currently trust as the category authority, which is who you are actually competing against in AI search.

Check 5: Confirm AI crawlers can reach you#

Check robots.txt for explicit allows on GPTBot, ClaudeBot, PerplexityBot, Google-Extended, and Applebot-Extended, then confirm an llms.txt file exists at your root. If crawlers are blocked or there is no llms.txt entry pointing to your best pages, the engines may never ingest the content that would have earned the citation.

Check 6: Inspect content structure for liftability#

AI engines quote short, direct, well-marked-up answers. Verify your key pages open each section with a 40 to 60 word answer capsule, use question-led headings, and carry schema (Organization, FAQPage, DefinedTerm). A page that buries its answer in paragraph six rarely gets lifted, no matter how authoritative it is.

What does a finished AI visibility audit look like?#

A finished audit is a single table: each buyer question as a row, each engine as a column, a cited/not-cited mark in every cell, plus the competitor cited and the fix assigned. The summary row is your per-engine citation rate. Here is the shape, using the real numbers from our chudi.dev sister-property panel (2026-06-12, 60 query-tests):

Engine	Questions tested	Cited	Citation rate	Top issue found
ChatGPT	20	7	35%	cites a mirror domain, not canonical
Perplexity	20	10	50%	strong; thin on comparison queries
Claude	20	10	50%	misses definitional queries
Overall	60	27	45%	ChatGPT canonicalization gap

The value is not the score. It is the named, ranked fix list that falls out of it: in this example, fix canonical signals so ChatGPT stops citing the mirror, then build the missing comparison and definitional pages. That is a roadmap, not a vanity metric.

Should you run an AI visibility audit yourself or buy one?#

Run it yourself first if you have 2 to 4 hours and only need a baseline; buy one when you need repeatability, more engines, or a defensible report for a stakeholder. The manual method is honest and free, and it teaches you exactly what the engines reward. The limit is time: re-testing 20 questions across 4 engines every month is the part worth automating.

Whichever path you choose, demand verifiable outputs. A credible AI visibility audit shows you the actual cited URLs and the per-engine breakdown. Be skeptical of any tool that hands you one blended number with no evidence underneath, because the evidence is the entire point. (The founder's background and methodology are on chudi.dev/about, if you want to vet the approach before trusting the numbers.)

Run your AI visibility audit#

If you have never tested it, the safest assumption is that at least one major engine is citing a competitor in answers your buyers read every week. You can run the 6 checks above by hand today, or get a forensic AI Visibility Audit that tests your buyer queries across every engine, names the domains cited in your place, and returns the ranked fix list. The number you get back is uncomfortable the first time and actionable every time after.

Have you ever actually asked ChatGPT a question your buyers ask, just to see whether it names you or your competitor?